Personalizing a context-free grammar using a dictation language model

• US 7,689,420 B2
• Filed: 04/06/2006
• Issued: 03/30/2010
• Est. Priority Date: 04/06/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Paek, Timothy S., Wu, Qiang, Chickering, David M., Badger, Eric Norman
• Primary Examiner(s): Dorvil; Richemond
• Assistant Examiner(s): He; Jialong
• Application Number:
  US11/278,899
• Publication Number:
  US 20070239454A1
• Time in Patent Office: 1,454 Days
• Field of Search: 704/9, 704/257, 704/250, 704/255
• US Class Current: 704/257
• CPC Class Codes: G10L 15/19 Grammatical context, e.g. disambiguation of the recognition hypotheses based on word sequence rules
• CPC Class Codes: G10L 2015/088 Word spotting

Abstract

Architecture for integrating and generating back-off grammars (BOG) in a speech recognition application for recognizing out-of-grammar (OOG) utterances and updating the context-free grammars (CFG) with the results. A parsing component identifies keywords and/or slots from user utterances and a grammar generation component adds filler tags before and/or after the keywords and slots to create new grammar rules. The BOG can be generated from these new grammar rules and can be used to process the OOG user utterances. By processing the OOG user utterances through the BOG, the architecture can recognize and perform the intended task on behalf of the user.

Claims

1. A system for generating a dictation language model in speech recognition applications, comprising:

a processor;

memory operatively coupled to the processor;

a detection component that identifies out-of-grammar (OOG) utterances, the OOG utterances are not recognized by context-free grammar (CFG) rules;

a parsing component configured to identify a slot in the OOG utterances;

a grammar generation component that generates a dictation language model by creating a new rule based in part on the OOG utterances, the grammar generation component configured to add a filler tag proximate to the slot to generate the new rule, wherein the slot is a dynamically adjustable list of text items, the slot references the rule to the list of text items, the slot being a direct object of the rule,

wherein the filler tag represents an acoustic element in the OOG utterances, and the filler tag combines with the slot to create the new rule;

a processing component that processes user utterances through the dictation language model before the CFG processes the utterances, the processing component further configured to communicate to a user all existing rules that contain the identified slot and receive from the user the user'"'"'s selection of one of the existing rules; and

a personalization component that updates old CFG rules with the new rule.

2. The system of claim 1, wherein the new rule is based in part on the OOG utterances and the old CFG rules, and wherein the personalization component is configured to eliminate utterances not frequently employed by the user from the CFG rules, and is further configured to specifically tailor the new rule to the user.

3. The system of claim 2, wherein only a portion of the CFG rules are re-written to generate the new rules.

4. The system of claim 1, wherein the filler tags are dictation tags.

5. The system of claim 1, wherein the new rules are created with filler tags that are based on words having a phonetic similarity to keywords.

6. A computer-implemented method for using a dictation language model to personalize a context-free grammar (CFG), comprising:

generating a dictation language model based in part on out-of-grammar (OOG) utterances in a computing device employing a processor operatively coupled to memory;

processing the OOG utterances on the computing device with the dictation language model before the CFG processes the OOG utterances;

verifying if the OOG utterances contain a slot that distinctly identifies an intended CFG rule to be applied to process the OOG utterances, identifying the slot in the OOG utterances, wherein the slot is a dynamically adjustable list of text items that reference the CFG rule to the text items;

tailoring the CFG rule for a user individually, the tailoring comprising;

communicating to the user all existing rules that contain the identified slot and receiving from the user the user'"'"'s selection of one of the existing rules;

requesting permission from the user to update the CFG with phrases based in part on the OOG utterances, the requesting comprising asking the user whether to add various parts of dictation text to the CFG, in various positions within the CFG, to create a new grammar rule;

updating the CFG with the phrases based in part on the OOG utterances if the user responds in the affirmative and not updating the CFG with the phrases based in part on the OOG utterances if the user responds in the negative;

identifying OOG utterances frequently used by the user;

determining if the OOG utterances frequently used by the user should be incorporated into the CFG;

updating the CFG with the OOG utterances frequently used by the user; and

eliminating utterances not frequently employed by the user from the CFG.

7. The method of claim 6, further comprising employing dictation tags to determine which rule is most likely an intended rule.

8. The method of claim 6, further comprising deriving a measure of phonetic similarity between the OOG utterance and keywords.

9. The method of claim 6, further comprising automatically updating the CFG with phrases based in part on the OOG utterances.

10. The method of claim 6, further comprising utilizing a predictive user model to determine the most likely command result based on the OOG utterances.

11. The method of claim 10, further comprising engaging in a confirmation of a predicted command result with the user.

12. The method of claim 11, wherein if the confirmation is correct, then automatically adding the command result to the CFG.

13. The method of claim 6, further comprising augmenting the CFG with utterances that the user does utilize.

14. A computer-implemented system for generating a dictation language model in command-and-control speech recognition applications, comprising:

computer-implemented means for identifying out-of-grammar (OOG) utterances that are not recognized by context-free grammar (CFG) rules;

computer-implemented means for generating new rules based in part on the OOG utterances, the means for generating new rules includes parsing the OOG utterances to identify keywords and slots, the slots are dynamically adjustable items of text lists and the keywords are text that allow a command or slot to distinguish from other commands or slots, the new rules include adding filler tags before or after the parsed keywords and parsed slots, the new rules are based in part on phonetic matching of keywords;

computer-implemented means for communicating to a user all existing rules that contain the identified slot and receiving from the user the user'"'"'s selection of one of the existing rules;

computer-implemented means for requesting permission from the user to update the CFG rules with phrases based in part on the OOG utterances, the requesting comprising asking the user whether to add various parts of dictation text to the CFG, in various positions within the CFG, to create new grammar rules;

computer-implemented means for updating the CFG rules with the phrases based in part on the OOG utterances if the user responds in the affirmative and not updating the CFG rules with the phrases based in part on the OOG utterances if the user responds in the negative;

computer-implemented means for processing the OOG utterances with the new rules before processing by the CFG rules;

computer-implemented means for updating the CFG rules with the new rules;

computer-implemented means for tailoring the CFG rules for the user individually; and

computer-implemented means for eliminating old rules from the CFG not frequently used by the user.

15. The computer-implemented system of claim 14, the new rules comprising a dictation grammar model generated in whole or in part.

16. The computer-implemented system of claim 14, the filler tags including dictation tags or garbage tags.